XYZ 12bit test chart

[marcogiustini]

Carl

very much, thanks.

In fact I could not find any info on the 'im' command but the extra info on the pixel function massively helps.

So, if I understand correctly, to create a full white patch (non DCI white) it would need three commands such as

im[y][x][0] = pixel(4095)
im[y][x][1] = pixel(4095)
im[y][x][2] = pixel(4095)

I guess I can create a simple loop for z = 0 to 2 and replace the component number with z?

Carsten
Thanks for investigating for me, appreciated. I'll report back if they ever come back to me from GIMP.

Thanks both!

[carl]

im is an 3D array created by the line

Code: Select all

im = numpy.zeros((height, width, 3), dtype=numpy.uint16)

So you're right, maximum XYZ in all pixels would be something like (not tested)

Code: Select all

for x in range(0, 1998):
  for y in range(0, 1080):
    for z in range(0, 3):
      im[y][x][z] = pixel(4095)

[marcogiustini]

that's very clear, thanks Carl

I think that's all I need. I'll experiment on Friday and report back.

[marcogiustini]

Hi all,

I've managed to create my charts and I checked them against the FS_Peak white one but I reckon my colour meter is not accurate enough to allow me to really evaluate this.

The code I used to make the DCI white is the following
(please forgive me if it's not the most efficient way of doing it, I am not a developer!)

Code: Select all

#!/usr/bin/python

from PIL import Image
import numpy
from libtiff import TIFF

width = 2048
height = 1080
filename = 'grey_10.tif'

im = numpy.zeros((height, width, 3), dtype=numpy.uint16)

# Convert 12 to 16-bit
def pixel(x):
    return x << 4


for x in range(0, width):
    for y in range(0, height):
			im[y][x][0] = pixel (3794)
			
        
for x in range(0, width):
    for y in range(0, height):
			im[y][x][1] = pixel (3960)
			
for x in range(0, width):
    for y in range(0, height):
			im[y][x][2] = pixel (3890)


tiff = TIFF.open(filename, mode='w')
tiff.write_image(im, write_rgb=True)
tiff.close()

link to grey-10.tiff: https://app.box.com/s/73iygotdctihf0mhp93kry6p9e34dl3u
link to CPL: https://app.box.com/s/8n9rrzj7uik3jdtbjvft9fxudjj9kl77

Is there a way to check the RGB values of the JPEG2000 frames inside the CPL? I know how to extract the frames but then what?

Thanks

[carl]

You can use this. Are you using Windows?

[marcogiustini]

that's great, thanks!

Yes, I'm a Windows user. Do you happen to have a compiled version? Otherwise I'll give it a go.

Thanks
Marco

[carl]

I've put an untested one up here -win32 for Windows 32 or -win64 for Windows 64.

[marcogiustini]

Thanks Carl,

I knew you would come to the rescue!
I've quickly tested my grey-10 chart and values are incorrect. X-3777 Y-3954 Z-3873
Also, values are - just by one point - changing over the row.

Code: Select all

3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3776 3954 3872
3776 3953 3872
3776 3953 3872
3776 3953 3872
3776 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873
3777 3954 3873

The official FS_White (at least the one I have) I've been using for a while reads 3795 3960 3891. It's off by only one digit.

Another FS_White from a respectable mastering facility reads slightly fluctuating values

Code: Select all

3787 3960 3884
3787 3961 3884
3786 3961 3885
3786 3961 3885
3787 3960 3885
3787 3960 3885
3787 3960 3884
3788 3960 3884
3788 3960 3884
3789 3959 3884
3789 3959 3884
3790 3960 3885
3790 3960 3885
3791 3960 3886
3792 3960 3887
3792 3961 3888
3793 3961 3888
3793 3961 3889
3794 3961 3889
3794 3961 3890
3794 3961 3890
3794 3961 3890
3794 3961 3890
3794 3961 3890
3794 3961 3889
3794 3961 3889
3794 3961 3889
3794 3961 3889

The TIFF I made reads - on Photoshop - 30352 - 31680 - 31120.

I believe I'd be happy with a discrepancy of just one but my chart seems to be off by 20+.

I then tried OpenDCP (OpenJPEG). Using the same TIFF file the outcome seems to be much closer to what it should be. And it's more consistent.

Code: Select all

3796 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3795 3961 3891
3795 3961 3891
3795 3961 3891
3795 3961 3891
3795 3961 3891
3795 3961 3891
3795 3961 3891
3796 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3795 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3796 3962 3892
3795 3962 3892
3795 3961 3891

What is your thought on the matter? I appreciate that DCP-o-MATIC is not supposed to deliver the level of accuracy I am looking for - that's why you can buy very expensive software for that - I am not asking you to 'fix' this, it's more for conversation's sake.

I am also surprised that the chart we have been using to set colours and brightness is not 100% accurate.

Thank you a lot for the dumpXYZ, that is really helping!

Let me know if I can provide more data.

Thanks
Marco

[carl]

What is your thought on the matter?

If you can give me your TIFF I can investigate.

I appreciate that DCP-o-MATIC is not supposed to deliver the level of accuracy I am looking for

Yes it is!

[marcogiustini]

Hi Carl,

Link to the TIFF: https://app.box.com/s/73iygotdctihf0mhp93kry6p9e34dl3u

DCP is also linked a few posts above.

Good to know that I can help you fine tuning the software then - I didn't want to look like the guy who's chasing ghosts!

Thanks